From Ukraine’s Energy System to Global Data Centres: Serhii Tokarev on the Prospects for Distributed Generation
The world’s most expensive scrap metal is currently sitting in the warehouses of major technology companies. These are latest-generation graphics processors worth billions of dollars, with nowhere to plug them in. While analysts and investors debate semiconductor shortages and the results of artificial intelligence model testing, technology giants have quietly encountered a more serious obstacle – a lack of electricity. Serhii Tokarev, investor and founder of Tokarev Foundation, explained what steps major companies are currently taking to address this problem and how Ukraine’s experience could help them.
Why Access to Electricity Is Becoming a Key Factor in Attracting AI Investment
New data centres require increasing amounts of electricity, connection equipment and grid capacity. According to Microsoft, the shortage of available power capacity will persist until at least the end of 2026. The next bottleneck in AI development will be the time needed to turn megawatts into operational computing capacity.
The International Energy Agency estimates that global digital infrastructure will require around 945 TWh of electricity in 2030, more than twice the amount consumed in 2024. Demand is expected to grow by 15% annually.
Power grids are already struggling to keep pace with this demand. In the US, waiting times for large transformers have increased from 12-16 weeks to two years. Microsoft CEO Satya Nadella has noted that chip shortages are no longer the main constraint on the development of AI infrastructure. Electricity has now become the critical issue.
“Europe is facing a similar problem. European countries plan to establish up to seven AI Gigafactories and attract more than €30 billion in funding to triple computing capacity by 2032. However, AI needs electricity today,” says Serhii Tokarev, founder of Tokarev Foundation.
The problem also concerns power density. A modern NVIDIA GB200 system consumes 120-132 kW, whereas a single server rack previously required only 7-12 kW. Future generations could consume up to 600 kW.
Despite these challenges, technology companies are already investing in power generation:
- Google has reached an agreement with Kairos Power on small modular reactors.
- Microsoft is financing the restart of the Three Mile Island reactor.
- Chevron, Joulent and GE Vernova plan to develop power generation capacity for a Microsoft data centre in Texas.
What Can Ukraine’s Experience with Distributed Generation Offer?
Distributed generation is one possible solution, and Ukraine provides a valuable example. Repeated attacks on its energy system during Russia’s full-scale invasion forced the country to find solutions quickly. As a result, Ukraine began deploying local sources of electricity.
“The government has reported the installation of around 600 small gas-fired generation units. Together, they provide more than 570 MW of capacity. Some are located near hospitals and water supply facilities. Ukraine already has 1.8 GW of distributed generation capacity, with another 1.5 GW planned,” noted Serhii Tokarev.
Ukraine’s experience cannot simply be transferred to the global level. However, it demonstrates how to respond when centralised grid capacity is insufficient. The country has already begun simplifying grid connection and permitting procedures. For companies developing artificial intelligence infrastructure, choosing suitable sites for data centres is becoming increasingly challenging. On-site or local generation, backup power systems and energy storage are becoming essential components of the energy infrastructure.
Artificial intelligence begins with code, but its development increasingly depends on megawatts. Countries competing for AI investment are now also focusing on the speed of grid connections and access to power generation capacity.



